Hello, I'm running two models in Mplus, one with a variable functioning as a mediator and one without. The mediation model looks like:
v6 ON v5; v5 ON v4 v2; v4 ON v3 v1; v3 ON v2 v1; v2 ON v1;
The non-mediation model looks like:
v6 ON v5; v5 ON v4 v2; v3 ON v1; v4 ON v1; v2 ON v1;
So, v3 no longer mediates between v1 & v2, and v4. It's still a predicted outcome of v1, but has no association with v2 or v4. All analysis properties are exactly the same between the two models.
The original model had 7 df. Since I deleted two paths, should I not have 9 df now? I was reading elsewhere on the forum about differences between Mplus df calculation and some other software, but wanted to make sure I wasn't making a different error (misunderstanding the model construction, meeting a limitation of the demo version, etc.). If this is a calculation difference, is there a reference that I might use?
You should obtain the same degrees of freedom with Mplus and Amos or any other software if you have the same model and the same estimator. Check to be certain that the same parameters are being estimated. It may be that there is a default you are not taking into account.
There is one parameter in the Mplus model that is not in the Amos model. This is the residual covariance between su_int and obcs_bs. This is the default in Mplus. This is why the degrees of freedom are not as you expected.